Add a basic YAML processor. Much of the language is still unimplemented.

lib/yaml/parser.dart

Index: lib/yaml/parser.dart
diff --git a/lib/yaml/parser.dart b/lib/yaml/parser.dart
new file mode 100644
index 0000000000000000000000000000000000000000..97726d21985148b9f142d022fb4f72b976bbcdbb
--- /dev/null
+++ b/lib/yaml/parser.dart
@@ -0,0 +1,959 @@
+// Copyright (c) 2012, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+/**
+ * Translates a string of characters into a YAML serialization tree.
+ *
+ * This parser is designed to closely follow the spec. All productions in the
+ * spec are numbered, and the corresponding methods in the parser have the same
+ * numbers. This is certainly not the most efficient way of parsing YAML, but it
+ * is the easiest to write and read in the context of the spec.
+ *
+ * Methods corresponding to productions are also named as in the spec,
+ * translating the name of the method (although not the annotation characters)
+ * into camel-case for dart style.. For example, the spec has a production named
+ * `nb-ns-plain-in-line`, and the method implementing it is named
+ * `_nb_ns_plainInLine`. The exception to that rule is methods that just
+ * recognize character classes; these are named `_is*`.
+ */
+class _Parser {
+  static final TAB = 0x9;
+  static final LF = 0xA;
+  static final CR = 0xD;
+  static final SP = 0x20;
+  static final TILDE = 0x7E;
+  static final NEL = 0x85;
+  static final HYPHEN = 0x2D;
+  static final QUESTION_MARK = 0x3F;
+  static final COLON = 0x3A;
+  static final COMMA = 0x2C;
+  static final LEFT_BRACKET = 0x5B;
+  static final RIGHT_BRACKET = 0x5D;
+  static final LEFT_BRACE = 0x7B;
+  static final RIGHT_BRACE = 0x7D;
+  static final HASH = 0x23;
+  static final AMPERSAND = 0x26;
+  static final ASTERISK = 0x2A;
+  static final EXCLAMATION = 0x21;
+  static final VERTICAL_BAR = 0x7C;
+  static final GREATER_THAN = 0x3E;
+  static final SINGLE_QUOTE = 0x27;
+  static final DOUBLE_QUOTE = 0x22;
+  static final PERCENT = 0x25;
+  static final AT = 0x40;
+  static final GRAVE_ACCENT = 0x60;
+
+  static final NULL = 0x0;
+  static final BELL = 0x7;
+  static final BACKSPACE = 0x8;
+  static final VERTICAL_TAB = 0xB;
+  static final FORM_FEED = 0xC;
+  static final ESCAPE = 0x1B;
+  static final BACKSLASH = 0x5C;
+  static final NBSP = 0xA0;
+  static final LINE_SEPARATOR = 0x2028;
+  static final PARAGRAPH_SEPARATOR = 0x2029;
+
+  static final _C_SEQUENCE_ENTRY = 4;

[Bob Nystrom, 2012/04/20 20:28:55]

Why make these private?

[nweiz, 2012/04/23 23:06:33]

Changed as part of the great de-underscoring.

+  static final _C_MAPPING_KEY = 5;
+  static final _C_MAPPING_VALUE = 6;
+  static final _C_COLLECT_ENTRY = 7;
+  static final _C_SEQUENCE_START = 8;
+  static final _C_SEQUENCE_END = 9;
+  static final _C_MAPPING_START = 10;
+  static final _C_MAPPING_END = 11;
+  static final _C_COMMENT = 12;
+  static final _C_ANCHOR = 13;
+  static final _C_ALIAS = 14;
+  static final _C_TAG = 15;
+  static final _C_LITERAL = 16;
+  static final _C_FOLDED = 17;
+  static final _C_SINGLE_QUOTE = 18;
+  static final _C_DOUBLE_QUOTE = 19;
+  static final _C_DIRECTIVE = 20;
+  static final _C_RESERVED = 21;
+
+  static final BLOCK_OUT = 0;
+  static final BLOCK_IN = 1;
+  static final FLOW_OUT = 2;
+  static final FLOW_IN = 3;
+  static final BLOCK_KEY = 4;
+  static final FLOW_KEY = 5;
+
+  /** The source string being parsed. */
+  final String _s;
+
+  /** The current position in the source string. */
+  int _pos = 0;
+
+  /** The length of the string being parsed. */
+  final int _len;
+
+  /** The current (0-based) line in the source string. */
+  int _line = 0;
+
+  /** The curent (0-based) column in the source string. */

[Bob Nystrom, 2012/04/20 20:28:55]

"current"

[nweiz, 2012/04/23 23:06:33]

Done.

+  int _column = 0;
+
+  /**
+   * Whether we're parsing a bare document (that is, one that doesn't begin with
+   * `---`). Bare documents don't allow `%` immediately following newlines.
+   */
+  bool _inBareDocument = false;
+
+  /**
+   * The line number of the farthest position that has been parsed successfully
+   * before backtracking. Used for error reporting.
+   */
+  int _farthestLine = 0;
+
+  /**
+   * The column number of the farthest position that has been parsed
+   * successfully before backtracking. Used for error reporting.
+   */
+  int _farthestColumn = 0;
+
+  /**
+   * The name of the context of the farthest position that has been parsed
+   * successfully before backtracking. Used for error reporting.
+   */
+  String _farthestContext = "document";
+
+  /** A stack of the names of parse contexts. Used for error reporting. */
+  List<String> _contextStack;
+
+  _Parser(String s) : _s = s, _len = s.length,

[Bob Nystrom, 2012/04/20 20:28:55]

I would format like:

_Parser(...)
  : _s = s,
    _len = ...

[nweiz, 2012/04/23 23:06:33]

Done.

+                      _contextStack = <String>["document"];
+
+  /**
+   * Return the character at the current position, then move that position
+   * forward one character. Also updates the current line and column numbers.
+   */
+  int _next() {
+    if (_pos == _len) return -1;
+    var char = _s.charCodeAt(_pos++);
+    if (_isBreak(char)) {
+      _line++;
+      _column = 0;
+    } else {
+      _column++;
+    }
+
+    if (_farthestLine < _line) {
+      _farthestLine = _line;
+      _farthestColumn = _column;
+      _farthestContext = _contextStack.last();
+    } else if (_farthestLine == _line && _farthestColumn < _column) {
+      _farthestColumn = _column;
+      _farthestContext = _contextStack.last();
+    }
+
+    return char;
+  }
+
+  /**
+   * Returns the character at the current position, or the character [i]
+   * characters after the current position.
+   *
+   * Returns -1 if this would return a character after the end or before the
+   * beginning of the input string.
+   */
+  int _peek([int i = 0]) {
+    var pos = _pos + i;
+    return (pos >= _len || pos < 0) ? -1 : _s.charCodeAt(pos);
+  }
+
+  /**
+   * The truthiness operator. Returns false if [obj] is null or false, true

[Bob Nystrom, 2012/04/20 20:28:55]

If you want to be fancy, consider code-formatting literals like true and null in
doc comments:  

Returns `false` if [obj] is `null` or `false`, `true`

[nweiz, 2012/04/23 23:06:33]

Done.

+   * otherwise.
+   */
+  bool _(obj) => obj != null && obj != false;
+
+  /**
+   * Consumes the current character if it matches [matcher]. Returns the result
+   * of [matcher].
+   */
+  bool _consume(bool matcher(int)) {
+    if (matcher(_peek())) {
+      _next();
+      return true;
+    }
+    return false;
+  }
+
+  /**
+   * Calls [consumer] until it returns a falsey value. Returns a list of all
+   * truthy return values of [consumer], or null if it didn't consume anything.
+   *
+   * Conceptually, repeats a production one or more times.
+   */
+  List _oneOrMore(consumer()) {
+    var first = consumer();
+    if (!_(first)) return null;
+    var out = [first];
+    while (true) {
+      var el = consumer();
+      if (!_(el)) return out;
+      out.add(el);
+    }
+  }
+
+  /**
+   * Calls [consumer] until it returns a falsey value. Returns a list of all
+   * truthy return values of [consumer], or the empty list if it didn't consume
+   * anything.
+   *
+   * Conceptually, repeats a production any number of times.
+   */
+  List _zeroOrMore(consumer()) {
+    var out = [];
+    var pos = _pos;
+    while (true) {
+      var el = consumer();
+      if (!_(el) || pos == _pos) return out;
+      pos = _pos;
+      out.add(el);
+    }
+  }
+
+  /**
+   * Just calls [consumer] and returns its result. Used to make it explicit that a

[Bob Nystrom, 2012/04/20 20:28:55]

Long line.

[nweiz, 2012/04/23 23:06:33]

Done.

+   * production is intended to be optional.
+   */
+  _zeroOrOne(consumer()) => consumer();
+
+  /**
+   * Calls each function in [consumers] until one returns a truthy value, then
+   * returns that.
+   */
+  _or(List<Function> consumers) {
+    for (var c in consumers) {
+      var res = c();
+      if (_(res)) return res;
+    }
+    return null;
+  }
+
+  /**
+   * Calls [consumer] and returns its result, but rolls back the parser state if
+   * [consumer] returns a falsey value.
+   */
+  _group(consumer()) {

[Bob Nystrom, 2012/04/20 20:28:55]

This name isn't clear to me. Maybe "attempt"?

[nweiz, 2012/04/23 23:06:33]

Changed to "transaction"; slightly longer, but much clearer.

+    int pos = _pos, line = _line, column = _column;
+    var res = consumer();
+    if (_(res)) return res;
+
+    _pos = pos;
+    _line = line;
+    _column = column;
+    return res;
+  }
+
+  /**
+   * Consumes [n] characters matching [matcher], or none if there isn't a
+   * complete match. The first argument to [matcher] is the character code, the
+   * second is the index (from 0 to [n] - 1).
+   *
+   * Returns whether or not the characters were consumed.
+   */
+  bool _nAtOnce(int n, bool matcher(int, int)) => _group(() {

[Bob Nystrom, 2012/04/20 20:28:55]

That doesn't do what you think it does. It declares that matcher takes two
Dynamic parameters whose names both happen to be "int".

[nweiz, 2012/04/23 23:06:33]

Fixed.

+    for (int i = 0; i < n; i++) {
+      if (!_consume((c) => matcher(c, i))) return false;
+    }
+    return true;
+  });
+
+  /**
+   * Consumes the exact characters in [str], or nothing.
+   *
+   * Returns whether or not the string was consumed.
+   */
+  bool _rawString(String str) =>
+    _nAtOnce(str.length, (c, i) => str.charCodeAt(i) == c);
+
+  /**
+   * Consumes and returns a string of characters matching [matcher], or null if
+   * there are no such characters.
+   */
+  String _stringOf(bool matcher(int)) =>
+    _captureString(_oneOrMore(() => _consume(matcher)));
+
+  /**
+   * Calls [consumer] and returns the string that was consumed while doing so,
+   * or null if [consumer] returned a falsey value. Automatically wraps
+   * [consumer] in `_group`.
+   */
+  String _captureString(consumer()) {
+    int start = _pos;
+    var res = _group(consumer);
+    if (!_(res)) return null;
+    return _s.substring(start, _pos);
+  }
+
+  /**
+   * Adds a tag and an anchor to [node], if they're defined. [props] should be a
+   * two-element list where the first element is a Tag or null and the second is
+   * a String or null.
+   */
+  _Node _addProps(_Node node, List props) {

[Bob Nystrom, 2012/04/20 20:28:55]

Using a List here is kinda gross. As tedious as it is, maybe make a little Tag,
String pair type.

[nweiz, 2012/04/23 23:06:33]

Done.

+    if (_(props[0])) node.tag = props[0];
+    if (_(props[1])) node.anchor = props[1];
+    return node;
+  }
+
+  /** Creates a MappingNode from [pairs]. */
+  _MappingNode _map(List<List<_Node>> pairs) {

[Bob Nystrom, 2012/04/20 20:28:55]

A Pair<A, B> class would help here. The code that is called by this would I
think be definitely clearer if its annotations were returning pairs (or better,
something semantically meaningful) instead of just lists.

[nweiz, 2012/04/23 23:06:33]

Done.

+    var content = new Map<_Node, _Node>();
+    pairs.forEach((pair) { content[pair[0]] = pair[1]; });

[Bob Nystrom, 2012/04/20 20:28:55]

=>?

[nweiz, 2012/04/23 23:06:33]

Doesn't work with "=", because it's a statement.

[Bob Nystrom, 2012/04/24 00:50:36]

? Assignment is an expression.

[nweiz, 2012/04/25 00:26:29]

Huh, I could have sworn this didn't work last time I checked. It works now,
though, so done.

+    return new _MappingNode("?", content);
+  }
+
+  /** Runs [fn] in a context named [name]. Used for error reporting. */
+  _context(String name, fn()) {
+    try {
+      _contextStack.add(name);
+      return fn();
+    } finally {
+      var popped = _contextStack.removeLast();
+      assert(popped == name);
+    }
+  }
+
+  /** Throws an error with additional context information. */
+  _error(String message) {
+    // Line and column should be one-based

[Bob Nystrom, 2012/04/20 20:28:55]

"."

[nweiz, 2012/04/23 23:06:33]

Done.

+    throw new SyntaxError(_line + 1, _column + 1,
+        "$message (in $_farthestContext)");
+  }
+
+  /**
+   * If [result] is falsey, throws an error saying that [expected] was
+   * expected.
+   */
+  _expect(result, String expected) {
+    if (_(result)) return result;
+    _error("expected $expected");
+  }
+
+  /**
+   * Throws an error saying that the parse failed. Uses `_farthestLine`,

[Bob Nystrom, 2012/04/20 20:28:55]

[] instead of ``.

[nweiz, 2012/04/23 23:06:33]

Done. I didn't remember that [] worked for members as well as parameters.

+   * `_farthestColumn`, and `_farthestContext` to provide additional
+   * information.
+   */
+  _parseFailed() {
+    throw new SyntaxError(_farthestLine + 1, _farthestColumn + 1,
+        "invalid YAML in $_farthestContext");
+  }
+
+  /** Returns the number of spaces after the current position. */ 
+  int _detectIndentation() {

[Bob Nystrom, 2012/04/20 20:28:55]

"countIndentation"? I would expect "detect" to return a bool.

[nweiz, 2012/04/23 23:06:33]

Done.

+    var i = 0;
+    while (_peek(i) == SP) i++;
+    return i;
+  }
+
+  /** Returns whether the current position is at the beginning of a line. */
+  bool get _atStartOfLine() {

[Bob Nystrom, 2012/04/20 20:28:55]

=> _column == 0

?

[nweiz, 2012/04/23 23:06:33]

Good idea. I wrote this before I added _column.

+    var char = _peek(-1);
+    return char == -1 || _isBreak(char);
+  }
+
+  /** Returns whether the current position is at the end of the input. */
+  bool get _atEndOfFile() => _pos == _len;
+
+  /**
+   * Given an indicator character, returns the type of that indicator (or null
+   * if the indicator isn't found.
+   */
+  int _indicatorType(int char) {
+    switch (char) {
+    case HYPHEN: return _C_SEQUENCE_ENTRY;
+    case QUESTION_MARK: return _C_MAPPING_KEY;
+    case COLON: return _C_MAPPING_VALUE;
+    case COMMA: return _C_COLLECT_ENTRY;
+    case LEFT_BRACKET: return _C_SEQUENCE_START;
+    case RIGHT_BRACKET: return _C_SEQUENCE_END;
+    case LEFT_BRACE: return _C_MAPPING_START;
+    case RIGHT_BRACE: return _C_MAPPING_END;
+    case HASH: return _C_COMMENT;
+    case AMPERSAND: return _C_ANCHOR;
+    case ASTERISK: return _C_ALIAS;
+    case EXCLAMATION: return _C_TAG;
+    case VERTICAL_BAR: return _C_LITERAL;
+    case GREATER_THAN: return _C_FOLDED;
+    case SINGLE_QUOTE: return _C_SINGLE_QUOTE;
+    case DOUBLE_QUOTE: return _C_DOUBLE_QUOTE;
+    case PERCENT: return _C_DIRECTIVE;
+    case AT:
+    case GRAVE_ACCENT:
+      return _C_RESERVED;
+    default: return null;
+    }
+  }
+
+  // 1
+  bool _isPrintable(int char) {
+    return char == TAB || char == LF || char == CR ||
+      (char >= SP && char <= TILDE) || char == NEL ||
+      (char >= 0xA0 && char <= 0xD7FF) || (char >= 0xE000 && char <= 0xFFFD) ||
+      (char >= 0x10000 && char <= 0x10FFFF);

[Bob Nystrom, 2012/04/20 20:28:55]

Style nit, but I would probably do one clause per line.

[nweiz, 2012/04/23 23:06:33]

Done.

+  }
+
+  // 22
+  bool _c_indicator(int type) => _consume((c) => _indicatorType(c) == type);
+
+  // 23
+  bool _isFlowIndicator(int char) {
+    var indicator = _indicatorType(char);
+    return indicator == _C_COLLECT_ENTRY || indicator == _C_SEQUENCE_START ||
+      indicator == _C_SEQUENCE_END || indicator == _C_MAPPING_START ||
+      indicator == _C_MAPPING_END;
+  }
+
+  // 26
+  bool _isBreak(int char) => char == LF || char == CR;
+
+  // 27
+  bool _isNonBreak(int char) => _isPrintable(char) && !_isBreak(char);
+
+  // 30
+  bool _b_non_content() => _consume(_isBreak);
+
+  // 33
+  bool _isSpace(int char) => char == SP || char == TAB;
+
+  // 34
+  bool _isNonSpace(int char) => _isNonBreak(char) && !_isSpace(char);
+
+  // 63
+  bool _s_indent(int n) => _nAtOnce(n, (c, i) => c == SP);
+
+  // 66
+  bool _s_separateInLine() => _group(() =>
+    _(_oneOrMore(() => _consume(_isSpace))) || _atStartOfLine);
+
+  // 69
+  bool _s_flowLinePrefix(int n) {
+    if (!_s_indent(n)) return false;
+    _zeroOrOne(_s_separateInLine);
+    return true;
+  }
+
+  // 74
+  bool _s_flowFolded(int n) => false; // TODO(nweiz): implement

[Bob Nystrom, 2012/04/20 20:28:55]

Throw UnsupportedOperationException instead. Here and elsewhere.

[nweiz, 2012/04/23 23:06:33]

That would break parsing of supported structures completely, since these methods
are getting called.

[Bob Nystrom, 2012/04/24 00:50:36]

Oh, it's relying on transaction to backtrack here? Never mind then.

+
+  // 75
+  bool _c_nb_commentText() {
+    if (!_c_indicator(_C_COMMENT)) return false;
+    _zeroOrMore(() => _consume(_isNonBreak));
+    return true;
+  }
+
+  // 76
+  bool _b_comment() => _atEndOfFile || _b_non_content();
+
+  // 77
+  bool _s_b_comment() {
+    if (_s_separateInLine()) {
+      _zeroOrOne(_c_nb_commentText);
+    }
+    return _b_comment();
+  }
+
+  // 78
+  bool _l_comment() => _group(() {
+    if (!_s_separateInLine()) return false;
+    _zeroOrOne(_c_nb_commentText);
+    return _b_comment();
+  });
+
+  // 79
+  bool _s_l_comments() {
+    if (!_s_b_comment() && !_atStartOfLine) return false;
+    _zeroOrMore(_l_comment);
+    return true;
+  }
+
+  // 80
+  bool _s_separate(int n, int c) {
+    switch (c) {
+    case BLOCK_OUT:
+    case BLOCK_IN:
+    case FLOW_OUT:
+    case FLOW_IN:
+      return _s_separateLines(n);
+    case BLOCK_KEY:
+    case FLOW_KEY:
+      return _s_separateInLine();
+    }

[Bob Nystrom, 2012/04/20 20:28:55]

Switches without defaults feel strange to me, especially since we can't rely on
static typing and flow analysis to ensure some branch gets hit. Maybe add:

default: throw 'unreachable';

or something like that. Here and elsewhere.

[nweiz, 2012/04/23 23:06:33]

Done.

+  }
+
+  // 81
+  bool _s_separateLines(int n) {
+    return _group(() => _s_l_comments() && _s_flowLinePrefix(n)) ||

[Bob Nystrom, 2012/04/20 20:28:55]

=> _group(() ... )

[nweiz, 2012/04/23 23:06:33]

I can't get this formatted so that everything fits cleanly on two lines when
using => there.

+      _s_separateInLine();
+  }
+
+  // 82
+  bool _l_directive() => false; // TODO(nweiz): implement
+
+  // 96
+  List _c_ns_properties(int n, int c) {

[Bob Nystrom, 2012/04/20 20:28:55]

"n" isn't a helpful name. "indent"?

[nweiz, 2012/04/23 23:06:33]

Changed n -> indent and c -> ctx (not "context" to avoid conflicting with the
method).

+    var tag, anchor;
+    tag = _c_ns_tagProperty();
+    if (_(tag)) {
+      anchor = _group(() {
+        if (!_s_separate(n, c)) return null;
+        return _c_ns_anchorProperty();
+      });
+      return [tag, anchor];
+    }
+
+    anchor = _c_ns_anchorProperty();
+    if (_(anchor)) {
+      tag = _group(() {
+        if (!_s_separate(n, c)) return null;
+        return _c_ns_tagProperty();
+      });
+      return [tag, anchor];
+    }
+
+    return null;
+  }
+
+  // 97
+  _Tag _c_ns_tagProperty() => null; // TODO(nweiz): implement
+
+  // 101
+  String _c_ns_anchorProperty() => null; // TODO(nweiz): implement
+
+  // 102
+  bool _isAnchorChar(int char) => _isNonSpace(char) && !_isFlowIndicator(char);
+
+  // 103
+  String _ns_anchorName() => _captureString(_isAnchorChar);
+
+  // 104
+  _Node _c_ns_aliasNode() {
+    if (!_c_indicator(_C_ALIAS)) return null;
+    var name = _expect(_ns_anchorName(), 'anchor name');
+    return new _AliasNode(name);
+  }
+
+  // 105
+  _ScalarNode _e_scalar() => new _ScalarNode("?", content: "");
+
+  // 106
+  _ScalarNode _e_node() => _e_scalar();
+
+  // 126
+  bool _ns_plainFirst(int c) {
+    var char = _peek();
+    var indicator = _indicatorType(char);
+    if (indicator == _C_RESERVED) {
+      _error("reserved indicators can't start a plain scalar");
+    }
+    var match = (_isNonSpace(char) && indicator == null) ||
+      ((indicator == _C_MAPPING_KEY ||
+        indicator == _C_MAPPING_VALUE ||
+        indicator == _C_SEQUENCE_ENTRY) &&
+       _isPlainSafe(c, _peek(1)));
+
+    if (match) _next();
+    return match;
+  }
+
+  // 127
+  bool _isPlainSafe(int c, int char) {
+    switch (c) {
+    case FLOW_OUT:
+    case BLOCK_KEY:
+      // 128
+      return _isNonSpace(char);
+    case FLOW_IN:
+    case FLOW_KEY:
+      // 129
+      return _isNonSpace(char) && !_isFlowIndicator(char);
+    }
+  }
+
+  // 130
+  bool _ns_plainChar(int c) {
+    var char = _peek();
+    var indicator = _indicatorType(char);
+    var match = (_isPlainSafe(c, char) && indicator != _C_MAPPING_VALUE &&

[Bob Nystrom, 2012/04/20 20:28:55]

This is kind of hard to understand. Maybe break it into separate checks?

[nweiz, 2012/04/23 23:06:33]

Done.

+        indicator != _C_COMMENT) ||
+      (_isNonSpace(_peek(-1)) && indicator == _C_COMMENT) ||
+      (indicator == _C_MAPPING_VALUE && _isPlainSafe(c, _peek(1)));
+
+    if (match) _next();
+    return match;
+  }
+
+  // 131
+  String _ns_plain(int n, int c) => _context('plain scalar', () {
+    switch (c) {
+    case FLOW_OUT:
+    case FLOW_IN:
+      return _ns_plainMultiLine(n, c);
+    case BLOCK_KEY:
+    case FLOW_KEY:
+      return _ns_plainOneLine(c);
+    }
+  });
+
+  // 132
+  void _nb_ns_plainInLine(int c) {
+    _zeroOrMore(() => _group(() {
+      _zeroOrMore(() => _consume(_isSpace));
+      return _ns_plainChar(c);
+    }));
+  }
+
+  // 133
+  String _ns_plainOneLine(int c) => _captureString(() {
+    if (_c_forbidden()) return false;
+    if (!_ns_plainFirst(c)) return false;
+    _nb_ns_plainInLine(c);
+    return true;
+  });
+
+  // 134
+  bool _s_ns_plainNextLine(int n, int c) => _group(() {
+    if (_c_forbidden()) return false;
+    if (!_s_flowFolded(n)) return false;
+    if (!_ns_plainChar(c)) return false;
+    _nb_ns_plainInLine(c);
+    return true;
+  });
+
+  // 135
+  String _ns_plainMultiLine(int n, int c) => _captureString(() {
+    if (!_(_ns_plainOneLine(c))) return false;
+    _zeroOrMore(() => _s_ns_plainNextLine(n, c));
+    return true;
+  });
+
+  // 154
+  _Node _c_s_implicitYamlKey(int c) => _group(() {
+    /* Indentation parameter is unused in this path */

[Bob Nystrom, 2012/04/20 20:28:55]

//

[nweiz, 2012/04/23 23:06:33]

Done.

+    var node = _ns_flowYamlNode(-10, c);

[Bob Nystrom, 2012/04/20 20:28:55]

What is -10 here?

[nweiz, 2012/04/23 23:06:33]

A value that's never referenced. Would null be better?

[Bob Nystrom, 2012/04/24 00:50:36]

Yeah, plus a comment explaining that.

[nweiz, 2012/04/25 00:26:29]

Done.

+    if (!_(node)) return null;
+    _zeroOrOne(_s_separateInLine);
+    return node;
+  });
+
+  // 155
+  _Node _c_s_implicitJsonKey(int c) => null; // TODO(nweiz): implement
+
+  // 156
+  _Node _ns_flowYamlContent(int n, int c) {
+    var str = _ns_plain(n, c);
+    if (!_(str)) return null;
+    return new _ScalarNode("?", content: str);
+  }
+
+  // 157
+  _Node _ns_flowJsonContent(int n, int c) => null; // TODO(nweiz): implement
+
+  // 158
+  _Node _ns_flowContent(int n, int c) => _or([
+    () => _ns_flowYamlContent(n, c),
+    () => _ns_flowJsonContent(n, c)
+  ]);
+
+  // 159
+  _Node _ns_flowYamlNode(int n, int c) => _or([
+    _c_ns_aliasNode,
+    () => _ns_flowYamlContent(n, c),
+    () {
+      var props = _c_ns_properties(n, c);
+      if (!_(props)) return null;
+      var node = _or([
+        () => _group(() {
+          if (!_s_separate(n, c)) return null;
+          return _ns_flowYamlContent(n, c);
+        }),
+        _e_scalar
+      ]);
+      return _addProps(node, props);
+    }
+  ]);
+
+  // 161
+  _Node _ns_flowNode(int n, int c) => _or([

[Bob Nystrom, 2012/04/20 20:28:55]

I'm OK with => for _group() method bodies which just wrap a single function
literal, but I don't like it for _or (or _and) where it contains a bunch of
arguments.

[nweiz, 2012/04/23 23:06:33]

I disagree. I can see how _or looks weird -- a list of functions isn't typical
in Dart -- but I think that's orthogonal to => vs {return}. I don't think moving
to curly braces adds any clarity here; the reader still has to learn what _or
means, but once they figure that out the more concise notation will be easier to
understand.

+    _c_ns_aliasNode,
+    () => _ns_flowContent(n, c),
+    () => _group(() {
+      var props = _c_ns_properties(n, c);
+      if (!_(props)) return null;
+      var node = _or([
+        () => _group(() => _s_separate(n, c) ? _ns_flowContent(n, c) : null),
+        _e_scalar]);
+      return _addProps(node, props);
+    })
+  ]);
+
+  // 170
+  _Node _c_l_literal(int n) => null; // TODO(nweiz); implement
+
+  // 174
+  _Node _c_l_folded(int n) => null; // TODO(nweiz); implement
+
+  // 183
+  _SequenceNode _l_blockSequence(int n) => _context('sequence', () {
+    var m = _detectIndentation() - n;
+    if (m <= 0) return null;
+
+    var content = _oneOrMore(() => _group(() {
+      if (!_s_indent(n + m)) return null;
+      return _c_l_blockSeqEntry(n + m);
+    }));
+    if (!_(content)) return null;
+
+    return new _SequenceNode("?", content);
+  });
+
+  // 184
+  _Node _c_l_blockSeqEntry(int n) => _group(() {
+    if (!_c_indicator(_C_SEQUENCE_ENTRY)) return null;
+    if (_isNonSpace(_peek())) return null;
+
+    return _s_l_blockIndented(n, BLOCK_IN);
+  });
+
+  // 185
+  _Node _s_l_blockIndented(int n, int c) {
+    var m = _detectIndentation();
+    return _or([() => _group(() {

[Bob Nystrom, 2012/04/20 20:28:55]

Wow this is hard to parse. Move () => _group... onto the next line and line up
it with }), and see if that helps.

[nweiz, 2012/04/23 23:06:33]

Done.

+        if (!_s_indent(m)) return null;
+        return _or([
+          () => _ns_l_compactSequence(n+1+m),
+          () => _ns_l_compactMapping(n+1+m)]);
+      }),
+      () => _s_l_blockNode(n, c),
+      () => _s_l_comments() ? _e_node() : null]);
+  }
+
+  // 186
+  _Node _ns_l_compactSequence(int n) => _context('sequence', () {
+    var first = _c_l_blockSeqEntry(n);
+    if (!_(first)) return null;
+
+    var content = _zeroOrMore(() => _group(() {
+        if (!_s_indent(n)) return null;
+        return _c_l_blockSeqEntry(n);
+      }));
+    content.insertRange(0, 1, first);
+
+    return new _SequenceNode("?", content);
+  });
+
+  // 187
+  _Node _l_blockMapping(int n) => _context('mapping', () {
+    var m = _detectIndentation() - n;
+    if (m <= 0) return null;
+
+    var pairs = _oneOrMore(() => _group(() {
+      if (!_s_indent(n + m)) return null;
+      return _ns_l_blockMapEntry(n + m);
+    }));
+    if (!_(pairs)) return null;
+
+    return _map(pairs);
+  });
+
+  // 188
+  List<_Node> _ns_l_blockMapEntry(int n) => _or([
+    () => _c_l_blockMapExplicitEntry(n),
+    () => _ns_l_blockMapImplicitEntry(n)
+  ]);
+
+  // 189
+  List<_Node> _c_l_blockMapExplicitEntry(int n) => null; // TODO(nweiz): implement
+
+  // 192
+  List<_Node> _ns_l_blockMapImplicitEntry(int n) => _group(() {
+    var key = _or([_ns_s_blockMapImplicitKey, _e_node]);
+    var value = _c_l_blockMapImplicitValue(n);
+    return _(value) ? [key, value] : null;
+  });
+
+  // 193
+  _Node _ns_s_blockMapImplicitKey() => _context('mapping key', () => _or([
+    () => _c_s_implicitJsonKey(BLOCK_KEY),
+    () => _c_s_implicitYamlKey(BLOCK_KEY)
+  ]));
+
+  // 194
+  _Node _c_l_blockMapImplicitValue(int n) => _context('mapping value', () =>
+    _group(() {
+      if (!_c_indicator(_C_MAPPING_VALUE)) return null;
+      return _or([
+        () => _s_l_blockNode(n, BLOCK_OUT),
+        () => _s_l_comments() ? _e_node() : null
+      ]);
+    }));
+
+  // 195
+  _Node _ns_l_compactMapping(int n) => _context('mapping', () {
+    var first = _ns_l_blockMapEntry(n);
+    if (!_(first)) return null;
+
+    var pairs = _zeroOrMore(() => _group(() {
+        if (!_s_indent(n)) return null;
+        return _ns_l_blockMapEntry(n);
+      }));
+    pairs.insertRange(0, 1, first);
+
+    return _map(pairs);
+  });
+
+  // 196
+  _Node _s_l_blockNode(int n, int c) =>
+    _or([() => _s_l_blockInBlock(n, c), () => _s_l_flowInBlock(n)]);
+
+  // 197
+  _Node _s_l_flowInBlock(int n) => _group(() {
+    if (!_s_separate(n+1, FLOW_OUT)) return null;

[Bob Nystrom, 2012/04/20 20:28:55]

Spaces around + here and elsewhere.

[nweiz, 2012/04/23 23:06:33]

Done.

+    var node = _ns_flowNode(n+1, FLOW_OUT);
+    if (!_(node)) return null;
+    if (!_s_l_comments()) return null;
+    return node;
+  });
+
+  // 198
+  _Node _s_l_blockInBlock(int n, int c) =>
+    _or([() => _s_l_blockScalar(n, c), () => _s_l_blockCollection(n, c)]);
+
+  // 199
+  _Node _s_l_blockScalar(int n, int c) => _group(() {
+    if (!_s_separate(n+1, c)) return null;
+    var props = _group(() {
+      var props = _c_ns_properties(n+1, c);
+      if (!_(props)) return null;
+      if (!_s_separate(n+1, c)) return null;
+      return props;
+    });
+    if (!_(props)) props = [null, null];
+
+    var node = _or([() => _c_l_literal(n), () => _c_l_folded(n)]);
+    if (!_(node)) return null;
+    return _addProps(node, props);
+  });
+
+  // 200
+  _Node _s_l_blockCollection(int n, int c) => _group(() {
+    var props = _group(() {
+      if (!_s_separate(n+1, c)) return null;
+      return _c_ns_properties(n+1, c);
+    });
+    if (!_(props)) props = [null, null];
+
+    if (!_s_l_comments()) return null;
+    return _or([
+      () => _l_blockSequence(_seqSpaces(n, c)),
+      () => _l_blockMapping(n)]);
+  });
+
+  // 201
+  int _seqSpaces(int n, int c) => c == BLOCK_OUT ? n - 1 : n;
+
+  // 202
+  void _l_documentPrefix() {
+    _zeroOrMore(_l_comment);
+  }
+
+  // 203
+  bool _c_directivesEnd() => _rawString("---");
+
+  // 204
+  bool _c_documentEnd() => _rawString("...");
+
+  // 205
+  bool _l_documentSuffix() => _group(() {
+    if (!_c_documentEnd()) return false;
+    return _s_l_comments();
+  });
+
+  // 206
+  bool _c_forbidden() {
+    if (!_inBareDocument || !_atStartOfLine) return false;
+    var forbidden = false;
+    _group(() {

[Bob Nystrom, 2012/04/20 20:28:55]

Couldn't you do:

return _group(() { ... });

and have the inner function return true or false instead of setting forbidden?

[nweiz, 2012/04/23 23:06:33]

No; that would move the cursor forward if it returned "true".

+      if (!_or([_c_directivesEnd, _c_documentEnd])) return;
+      var char = _peek();
+      forbidden = _isBreak(char) || _isWhite(char) || _atEndOfFile;
+      return;
+    });
+    return forbidden;
+  }
+
+  // 207
+  _Node _l_bareDocument() {
+    try {
+      _inBareDocument = true;
+      return _s_l_blockNode(-1, BLOCK_IN);
+    } finally {
+      _inBareDocument = false;
+    }
+  }
+
+  // 208
+  _Node _l_explicitDocument() {
+    if (!_c_directivesEnd()) return null;
+    var doc = _l_bareDocument();
+    if (_(doc)) return doc;
+
+    doc = _e_node();
+    _s_l_comments();
+    return doc;
+  }
+
+  // 209
+  _Node _l_directiveDocument() {
+    if (!_(_oneOrMore(_l_directive))) return null;
+    var doc = _l_explicitDocument();
+    if (doc != null) return doc;
+    _parseFailed();
+  }
+
+  // 210
+  _Node _l_anyDocument() =>
+    _or([_l_directiveDocument, _l_explicitDocument, _l_bareDocument]);
+
+  // 211
+  List<_Node> _l_yamlStream() {
+    var docs = [];
+    _zeroOrMore(_l_documentPrefix);
+    var first = _zeroOrOne(_l_anyDocument);
+    if (!_(first)) first = _e_node();
+    docs.add(first);
+
+    _zeroOrMore(() {
+      var doc;
+      if (_(_oneOrMore(_l_documentSuffix))) {
+        _zeroOrMore(_l_documentPrefix);
+        doc = _zeroOrOne(_l_anyDocument);
+      } else {
+        _zeroOrMore(_l_documentPrefix);
+        doc = _zeroOrOne(_l_explicitDocument);
+      }
+      if (_(doc)) docs.add(doc);
+      return doc;
+    });
+
+    if (!_atEndOfFile) _parseFailed();
+    return docs;
+  }
+}
+
+class SyntaxError extends Error {
+  final int _line;
+  final int _column;
+
+  SyntaxError(int this._line, int this._column, String _msg) : super(_msg);
+
+  String toString() => "Syntax error on line $_line, column $_column: $_msg";
+}